Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: 3 Questions Regarding the Cox Proportional Hazard Model


From   Maarten Buis <maartenlbuis@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: 3 Questions Regarding the Cox Proportional Hazard Model
Date   Fri, 13 Jan 2012 14:15:10 +0100

On Thu, Jan 12, 2012 at 7:33 PM, Yuval Arbel wrote:
> 1. Is there any way to calculate marginal elasticities between the
> hazard rate and each variable? apparently what I need  is something
> similar to the double-log model in simple regression analysis, or
> alternatively something similar to -dprobit-

The strength _and_ weakness of Cox regression is that it does not
estimate the baseline hazard. It is a strength since you cannot make
an error in the specification of the baseline hazard if you just do
not estimate it. It is a weakness in that it makes it hard to say
anything about the hazard or anything that depends on it, like
marginal effects or elasticities. The only effect size that really
makes sense in a Cox model is the hazard ratio. This is one of the
advantages of -stpm2- (see: -findit stpm2-), it estimates the baseline
hazard function, so now it makes sense to think about alternative
effect sizes.

> 2. Suppose, the PH assumption does not work in my sample. If my only
> concern is the coefficients of variables and their marginal effects on
> the survival rate, and not projected survival rates, does rejection of
> the PH assumption mean I cannot use the results at all without
> stratifications of the sample?

You cannot ignore the proportional hazard assumption, but
stratification is not the only solution. You can also model the
relationship between the hazard ratio and time using the -tvc()-
option. I have explained this to you before:
<http://www.stata.com/statalist/archive/2012-01/msg00282.html>.

> 3. If I run the -tvc- option, how do I make hypothesis testing on the
> interaction terms between time and the variables?

I gave you a link some time ago that included an example:
<http://www.stata.com/statalist/archive/2012-01/msg00211.html>.

-- Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index